| Keywords |
Filter; Remaining useful life; Fan differential pressure; Machine learning; Multinomial logistic; regression |
| Abstract |
This study proposes a method for predicting the remaining useful life (RUL) of air handling unit (AHU) filters without directly measuring the differential pressure across the filter. Instead, the approach utilizes operational data collected from the Building Automation System (BAS). By analyzing the supply fan’s differential pressure and the accumulated dust load, the filter’s RUL is estimated. A multinomial logistic regression algorithm is applied for model training, and demonstrates reliable accuracy across all classified RUL categories. The proposed method enables data-driven filter maintenance, improving both cost efficiency and indoor air quality. This approach is particularly valuable for buildings that lack dedicated filter pressure sensors. |